# Change image transformations during training

**URL:** <https://discuss.huggingface.co/t/change-image-transformations-during-training/139970>\
**Category:** 🤗AutoTrain\
**Created:** [February 10, 2025, 8:41am UTC](https://discuss.huggingface.co/t/change-image-transformations-during-training/139970 "2025-02-10T08:41:24Z")\
**Posts on this page:** 1\
**Page:** 1

<div class="post-metadata">

**Author:** ![cloudchief](https://avatars.discourse-cdn.com/v4/letter/c/b77776/32.png) [@cloudchief](https://discuss.huggingface.co/u/cloudchief)\
**Post date:** [February 10, 2025, 8:41am UTC](https://discuss.huggingface.co/t/change-image-transformations-during-training/139970/1 "2025-02-10T08:41:24Z")

</div>

Hi!

I want to use autotrain for image classification and object detection. There is a set of fixed transformation for augmentation in [image\_classification/utils.py](https://github.com/huggingface/autotrain-advanced/blob/b5c98fbe3aab61101e9c8f7f6c64407cbb68e400/src/autotrain/trainers/image_classification/utils.py)

I wonder what the best way would be to alter the transformations before training? There are cases where I’d really like to add others or change the existing ones. As far as I could see those aren’t configurable, are they?

What would be the best approach to alter the transformations?